Big Data Modelling of the Knowledge Economy
Robert B. Mellor
Computing and Mathematics
Kingston University
London, UK
In Tomé, E. (ed). Proceedings of the International Conference Theory and
Application in the Knowledge Economy (2016) ISBN 978‐989‐20‐6806‐0; p 30-43.
Abstract
A computer-generated 3D model is used to illustrate the advantages of in silico
techniques in virtual reality. The model is derived from theoretical data available for
SMEs in service industries and for example enables a business owner (or
consultant) to identify where the particular organization is on this three-dimensional
landscape and draw quantitative conclusions about fruitful future directions for the
organization, as well as how high the benefits will be and what costs will be due
along the journey.
Clearly this landscape map is of immense value for academics and practitioners
alike, and is available “ready-to-go” in a generic and easily-applicable form. This
Masterclass will go further and show participants the use of software tools to
construct a contour landscape that is specific for their own client organization.
We will create the basic 3-dimensional fold in Maple and discuss the implications for
SME growth and development. We will then move to Markov Chain Monte Carlo
modelling, which - put simply - is throwing virtual balls down the basic fold to show
how the fold can be used and manipulated to model and predict outcomes of
“Knowledge Engineering”. In particular we will model; adding multi-skilled innovators,
adding network input from the external environment, looking at the cost of controlling
management by showing how much bad information gatekeeping at departmental
manager level actually costs and investigating the value-adding effect of intellectual
capital.
If time allows, participants will be presented with a small case study and – together
with the class leader – will work through a simple solution thus illustrating the ease
and applicability of the model in a practical and hands-on fashion.
Keywords: Business Analysis, Computer model, Innovation, profitability, SME.
Introduction
The traditional resource-based view (RBV) of the firm arises from an industrial
organization approach that assumes there is a business reality about which
information can be gathered and processed to arrive at rational decisions which are
subsequently acted upon by management layers in order to realize the strategy
involved. The RBV however is very stable and baulks at the concept of fundamental
change. Thus a newer approach was born which puts change at the heart of the
argument; this Knowledge Based View (KBV) is centred around a firms dynamic
capability which sounds fine but there can be considerable confusion around what
that is and how to acquire and develop it, which in turn forms the rich ups-and-downs
playground of knowledge management theorists. Some industry areas have
emerged in the past 20 years which are more homogenous and thus amenable to
modelling, thus the Innovation Based View (IBV) has been postulated (Mellor,
2015a) to be especially relevant for these sectors, meaning that the IBV is an
application for some subsets of the KBV. Relevant sectors in particular involve SMEs
in service industries because they typically do not have many complicating factors
that make the baseline uneven, like foreign subsidiaries, research gaps, patent
barriers, factories and raw materials, production lines etc.
As it stands, KBV is about a firm’s aptitude to realize and deal with change, but is
very fuzzy about where that change is leading to, and even how to provoke the
‘correct’ change. Thus KBV theory does not deliver a good method to navigate
towards success as defined by entrepreneurial standards including firm survival and
improved performance in the face of change. However one improvement is that KBV
begins to add subjectivity to strategy by admitting that owners and managers can
make mistakes and indeed the term “core rigidity” is now almost as common as the
expression “core competencies”. None the less, innovation (in the sense of KBV)
management research reveals its deficiencies by still being largely unable to explain
precisely how Schumpeterian (i.e. transient) competitive advantage comes about,
and how this entrepreneurial effect could be better sustained.
Improved scientific methods, including control experiments, are needed to pin down
this important question but these techniques are rarely applicable in the fields of
Economics and Business Studies; indeed where such instances do occur they are
serendipitous and often oblique. This problem has led to an innovative laboratory
approach being developed in computer-generated realities and has the advantage
that experiments in silico are not only more rigorous but also cheaper and faster than
conventional approaches e.g. case studies. One such computer model has resulted
in a data-driven 3D model which completely explains the phases reported 40 years
ago that are characteristic of the development of SMEs (e.g. Grainer, 1972). The
computer model has also been used experimentally to predict outcomes associated
with structural changes within organizations, and preliminary results are encouraging
(Mellor 2014a, 2014b and 2015b). The present model is derived from theoretical
data (Mellor, 2011) and for example enables a business owner or consultant to
identify where the particular organization is on this 3D landscape and draw
quantitative conclusions about fruitful future directions for the organization, how high
the benefits will be and what costs will be due along the journey.
To test the hypotheses involved and to perfect this model we have chosen a “Big
Data” approach using statistically significant amounts of real data from a longitudinal
dataset housed at the Leibniz-Institut für Wirtschaftsforschung (Halle Institute for
Economic Research) in Germany. Using the German data in the UK computer model
we aim to present a rigorous and evidence-based unified model of business growth
and development.
This communication details the model and furthermore provides twelve practical
pointers of use to academics and practitioners alike.
Methodology
SMEs in service areas are normally unencumbered by such assets as production
lines, patents etc and compete by incremental innovation, a process of being just
ahead of the rest in some small way. This “mutual inspiration” is derived in many
ways, often by communication between individual of different backgrounds, or
outlooks, or educational specialization. Thus diversity adds to innovation and this
“Diversity Innovation” is thought to be the most important form of innovation in small
service industries because it provokes a valuable “mutual inspiration” effect. To
quantify this effect persons are represented as nodes (the number of people is
represented by P) and they are joined by ties, the number of links or ties between
nodes is the Diversity Innovation (DI) number. As the DI number increases the
potential for innovation increases. Clearly the DI number merely represents the
potential for the type of small-scale, incremental innovations thought of as being
important in small growing organizations (SMEs) and not the sum knowledge in the
organization. A simple equation has been published (Mellor 2011) that shows that
the larger the number of people in communication, the larger the probability that
some kind of “mutual inspiration” will occur. Clearly one could be tempted to put lots
of different people in a room and wait for an innovation to take form but in mercantile
organizations the transaction costs for this type of communication are too high. To
illustrate this, think of a company with 120 employees; the DI number = 7140, so
talking continuously and without any break, this would take 595 hours or 16 man-
weeks of working time just for employees to talk to each other for 5 minutes,
excluding that any employees got a chance to repeat any conversations (or do any
work). It would be expected that the total outlay even only in terms of salary for this
“speed-dating” exercise would exceed the value of any innovation arising from it.
Even worse, each further employee hired would take around 10 man-hours to talk to
existing employees for 5 minutes each, so the transaction costs are considerable.
There are also transaction costs involved in management, firstly that as departments
grow to size 50 employees, the transaction costs incurred by the manager mean that
the manager uses all their available time to communicate with them and that more
employees means too little time and a consequent reduction in communication
quality, thus organizations tend to enter into a policy of serial departmentalization
during growth when any department reaches maximum 50 employees. As the
organization grows, it is split into departments and this has consequences for the DI
number, as shown in figure 1 (below).
Figure 1: DI number (subjected to a smoothing algorithm) in a population
between 1 and 250 nodes (employees), splitting the organization into equally-
sized parts when each department reaches size 60.
To this 2D curve we can now add the third dimension. The third dimension of the
"landscape" is given by the J-curve (a hockey stick curve), because the J-curve
represents a transition where things get worse before they get better, this effect is
known from several disciplines, e.g. medicine, national economics and Business
Process Re-engineering. J-curves are known from national economic data (shown in
figure 2 in blue) but in this case data calculated from the Business Process
Reengineering literature is used (violet stippled in figure 2). The difference between
these curves is that bankrupt states can keep going, but bankrupt companies (in a
Capitalist democracy) cannot.
250 employees
Figure 2: The financial journey from low (left) to high (right) value for national
economic development (blue line) or companies undergoing re-organization
(stippled line).
On this axis there are no units for innovation (“innos”). However we can use %
instead on a 1-100% scale where the value is the amount derived from the maximum
(100%) possible. We can do this because the maximum possible, for an organization
of known size (number of employees) is known from the DI number. A finer
granularity can be obtained by benchmarking the degree of innovation the firm
involved against other firms in the same industry segment.
When the curve (for the business process reengineering data) from figure 2 is added
as the third dimension to figure 1 using Maple software (maplesoft.com), then the
following 3D fold is obtained:
Figure 3: Figure 3A, the basic fold and figure 3B illustrating the axis dependencies;
X axis as growth (number of employees), Y axis as value (e.g. annual turnover) and
Z axis as openness to innovation on a benchmarked scale.
Zero
Results from modelling in Maple.
From figure 3 any business owner (or consultant) can identify where the particular
client organization is on this 3D landscape and draw quantitative conclusions about
fruitful future directions for the organization. However the fold in Figure 3 does not
indicate the costs associated with any future plans. To estimate costs figure 2 is
revisited, because figure 2 shows that the J-curve slips below zero. In the analogy
(to a valley) presented here the zero level is thought of as being represented by a
lake (the ‘tarn’ of ‘knowledge valley’). Organizations progressing from a low-
innovation state to a high-innovation state, for example by the process known a
Business Process Reengineering, often transiently lose profitability. Using figures
from the relevant literature, Mellor (2011) derived two important factors:
1. That for each individual firm the drop in profitability can be greater or lesser,
but the maximum drop is around 40%, greater loss than this generally leads to
rapid downsizing and the shedding of knowledge assets and core
competencies which in turn leads to bankruptcy.
2. The time of transition is important, the shorter the time in the loss-making
zone the lesser the chance of negative consequences like bankruptcy.
The ‘tarn’ fits into the fold as shown in Figure 4.
Figure 4: The fold up to organizations of size 200 employees showing the position of
the zero profitability-level as a lake. Using the terminology of small business
research, companies just starting up are designated ‘mice’, those at the peak of
performance are ‘gazelles’ and less innovative medium-sized enterprises are
‘elephants’.
Danger point
Figure 4 indicates two further general lessons to ponder upon:
3. The second round of departmentalization cuts profitability back to near-zero
levels. Clearly while owners are growing their organization family or friends
may be installed as managers, but at the second round this becomes a
dangerous practice and consultants etc should be recommending that
competent professionals are hired.
4. The depth of the ‘tarn’ increases with size, so the costs for a medium-sized
organization to cross to the innovative side are more than the costs for a small
organization. So organizations that from the very beginning stay on the
innovative side avoid costs and danger later in their life.
Figure 4 also illustrates that around the zero level (on both sides) is the home of
social entrepreneurship, those organizations that aim at rarely or only transiently
making a profit and these may be churches, charities or similar. Organizations can
also exist beneath the zero level; these are for example state organizations. These
adopt a pendulum path from side to side, rarely surfacing above the zero level. For
example state bureaux exist in the depths of the tarn because the pull up to the right
of the J-curve towards efficiency and better public service (modernization & e-
government initiatives, etc) is balanced by a pull to autocracy and the left of the J-
curve (political expedients, the dominance of political targets, civil service “empire-
building”, etc), leaving them always loss-making at around the lowest point.
Furthermore if any state service does become (potentially) profit making then state
policies may well indicate that this portion is to be sold off to private enterprise,
leaving the loss-making remainder to sink down again to the mid-point.
It is known that unhindered knowledge flow and the rapid dissemination of
incremental innovations arising from this, is especially important in SMEs (Desouza
& Awazu, 2006. Serenko et al, 2007) and this aspect has been applied to studies on
management control in SMEs (for some examples see e.g. Ditillo, 2004). In terms of
management control, Williamson (1985) argues that "self-interest seeking with guile"
(Williamson, 1985 p30, Williamson 1993, p97) is inherent in human behaviour and
thus that it is difficult to identify trustworthy individuals in organizations. The logical
result of this is that it is necessary for organizations to structure themselves with
internal checks and balances in such a way as if individuals cannot be trusted, which
clearly adds a further layer to the internal transaction costs that organizations incur.
While this is very rational, it begs the question; how high is the cost of hindering
knowledge flow at departmental manager level? This is because it is inefficient to
have expensive policing if the cost of the crime is petty. Modelling can give an
indication of the cost because the links between departments can be discounted in
the algorithm. Figure 5 shows the effect of hindering knowledge flow at departmental
manager level.
A B
Figure 5: The fold with; figure 5A, open information gateways at departmental
manager level and figure 5B, closed information gateways at departmental manager
level. Note in addition to poorer performance overall in figure 5B, there is a further
performance dip (indicated) in figure 5B.
With all four inefficient department leaders and the organization at the peak of
performance, the point where the most difference would be found, overall
performance was down by a mere 6.4% (SD 0.05, n=10). As an across the board
rule and subject to some variation (within the constraints given in real life like the
vagrancies of exactly how poor the information gatekeeping is) then this (6.4%)
percentage of financial under-performance will apply to organizations generally
having four departments. Further modelling showed that one departmental manager
blocking information and knowledge flow to other departments reduced the financial
performance of an organization by 1.4% in the local department and a further 1.2%
spread across the remainder of the organization. Two such managers reduced
overall performance by 4.1%. Results indicate:
5. In high-innovation environments the costs of "self-interest seeking with guile"
are not sufficiently high to justify heavy preventative measures, at least in the
short-term. However, if the workforce attached to that departmental manager
becomes demotivated – a scenario that may be more common in low-
innovation work environments – then costs may increase significantly.
6. In the “closed gate” scenario a crisis occurred at the third round of
departmentalization at around size 150 employees which is not seen in open
gatekeeping. Perhaps this indicates e.g. the departmental managers
squabbling over the division of budget and assets as the organization grows
from three departments to four departments - costly disagreements.
Performance dip
Thus in addition to locating where an organization is on the knowledge valley fold
(figure 4) one can now draw conclusions about the amount of resources that should
be deployed to ensure the internal policing of the departmental managers behaviour.
Results from Markov Chain Monte Carlo modelling.
Although the fold shown in figure 3A is not a true Markov Chain, it can be used as
such for the purposes of modelling. Markov Chain Monte Carlo simulations (see e.g.
Chib and Greenberg, 1996 for a classical review of Markov Chain Monte Carlo
techniques or Robinson, 2014, for a more recent overview) in this case start by
rolling virtual balls from the ‘mouse’ origin down the fold. Everything being equal, the
balls will hit the far end of the J-curve in a random fashion i.e. there will be a
homogenous distribution of exit impacts along the curve. However the fold can also
be manipulated e.g. squeezed or attenuated, and the change in distribution of the
exit impacts visualized as a scatter plot. If a statistically significant correlation exists
between the factor being tested and adding value, then the scatter plot will form a
probability density function (a Gaussian distribution or ‘normal curve’). On a classical
Gauss curve the scale parameter (small sigma, σ) is 1.00. How much a given curve
varies from a classical Gaussian distribution can therefore be assessed by
measuring the scale parameter of the curve; the more it diverges from 1.00 the less
correlation there is. Values of scale parameter below 1.00 indicate a very sharp peak
and those over 1 indicate a spread peak (referred to as a platykurtic distribution), i.e.
of low correlation. Values of 3.0 and over indicate a correlation that, for practical
purposes, is so spread and weak that it can be ignored.
Workplace innovation/mutual inspiration is at its most simple - and powerful - when
communication problems do not exist, i.e. where the two or more specialists are
literally embodied within the same person. Thus one person – one human brain – is
able to look at a problem with the eyes of (say) a geologist and a businessperson, or
teacher, simultaneously. Indeed employees with a multi-specialised background i.e.
possessing ‘T-shaped’ or ‘A-shaped’ skills (Tsai and Huang, 2008; Mellor, 2005)
have been implicated in raised innovation. Furthermore they are likely to be highly
qualified and thus may be assumed to work in middle management. Accordingly the
second dimension (value) in the knowledge valley fold can be converted from
company annual turnover to salary, and a ‘knowledge trail’ (Mellor, 2011) spanning a
reasonably wide band of middle management salary inserted. This band thus forms
a bifurcating pathway along the fold and exits occur on both the left and right-hand
sides of the fold on the Z-axis. Along this bifurcating band, innovators are
represented not as one node, but as two overlapping nodes, i.e., that no tie is
needed to traverse between the nodes. This is done to represent the lack of
transaction costs for communication in this individual, so in this case the Monte Carlo
ball can progress two nodes down the fold without incurring ‘costs’. The results
obtained are shown in figure 6 and illustrate that low-innovation companies i.e. on
the left side of the J-curve) appear to be able to profitably use multi-specialised
individuals in middle management and indeed things may seem to get quite a lot
better (remember the vertical axis is value) as the organisation opens up to
innovation and change. However, as they proceed along the skewed Gauss curve
(moving right) they approach a decline, the nature of which could be very diverse
depending on branch and industry, but in any case may be disadvantageous.
Figure 6: Showing (Figure 6A) the relationship between placing innovators in a band
marked by the block, and exit impacts on the J-curve. Figures 6B and 6C; scatter
plot values derived from exit impacts on the both sides of the fold.
Figure 6 also shows that companies showing a high percent of use of innovation can
likewise profitably use multi-specialised individuals in middle management. The
model indicates that only low or no risk is associated with this strategy and, if it
works, things may get much better very rapidly indeed. After this gain in value is
realised, added value will reach a plateau and level off and thereafter increasing the
numbers/density of multi-skilled individuals (i.e. hiring more) should not have much
further effect. Thus:
7. Adding innovations/innovators to low-innovation organizations can have
positive effects relatively quickly (the “low-hanging fruit” effect) but the
temptation to add more and more innovation should be resisted unless a full-
scale business process reengineering operation is planned.
8. Adding innovations/innovators to high-innovation organizations is low-risk and
can have some positive effects but these are more likely to be reaching
performance gains more rapidly than overall performance improvement. One
way in which organizations can profit would be in cases where these
innovations are poorly aligned with ‘core competencies’ and the parent
organization can profitably spin-out and later sell these innovations
encapsulated in a start-up firm.
In the model used above, innovators were added into the bifurcated stream at
random. Clearly models can be produced where they are located more upstream or
more downstream. A number of innovators were used where the scale parameter
(small sigma, σ) was 1.5 and the experiment repeated with a Pareto distribution
upstream or downstream, the results are shown in Table 1.
80% upstream,
20% downstream
Random (control) 20% upstream,
80% downstream
High innovation
side
1.21 1.50 1.61
Low innovation
side
1.22 1.50 1.63
Table 1: The effect of the placement of innovators on the value of the scale
parameter (σ) in the resulting scatter plots (SD 0.05, n=10).
Table 1 shows that in both high- and low-innovation firms, a ‘tighter’ peak was
obtained when the innovators were placed upstream. This implies:
9. Putting innovations and innovators into an expanding organization is good,
but the effect of parachuting innovators onto an already existing and less
innovative middle management, is less effective than having started with an
innovative middle management right from the beginning.
In the modelling described above, a rather high number of innovators has been
used; to achieve a scale parameter (σ) of 1.5 in the scatter plots the number was
50% of all middle managers. Clearly this is rather a lot and at odds with everyday
reality. However innovations do not have to have an endogenous source i.e. arise in
situ from within the firm, they can arrive from outside the firm and be imported. To do
this one typically needs a network. Networks can be modelled, in this case Markov
Chain Monte Carlo modelling was performed as before but previously all Monte
Carlo balls started at the origin (the ‘mouse’ position in figure 4) of the fold and a
variable number of ‘innovators’ were placed in a band corresponding to a middle-
management salary as above. However in the experiments described in this section
packets of useful information could arrive from individuals outside of the
organisation. Thus in this experiment the same overall number of balls were rolled
down the knowledge valley fold as before, except that a variable proportion of these
balls appeared ‘spontaneously’ in a random fashion within the innovator band. For
example, 50:50 represents a situation where each innovator can get 1 datum of “just-
in-time” knowledge from his/her external network. Experiments tested the proportion
of balls at 33: 66 and at 50: 50 and again at 66: 33. This is meant to simulate the
number of innovations (number of balls) being constant while the number of
innovators responsible for them was varied, the balls appearing at random represent
inspiration coming in from outside the organisation (and thus can appear anywhere
along the band). The ratios represent one innovator using their network to harvest
two innovations (33.3: 66.6), one innovator bringing in one inspiration from outside
(50: 50) and finally the network value being one inspiration from outside for every
two innovators in the organisation (66.6: 33.3). The results are shown in table 2:
Value of scale parameter
One innovator and two randomly-arising
external innovations.
1.15
One innovator and one randomly-arising
external innovation.
1.41
Two innovators and one randomly-arising
external innovation.
1.49
Innovators only (control) 1.50
Table 2: The effect of changing the proportion of Monte Carlo balls starting at the
origin of the fold, on the scale parameter of the resulting scatter plot (SD 0.05, n=10).
The results of these experiments show:
10. The ratio of innovators to non-innovators in management can be as low as 1:6 if
working “just-in-time” knowledge networks are available.
11. The value of exogenous innovation when arriving in a timely and applicable
fashion appears to be only marginally less valuable than ‘home-made’
endogenous innovation arising in situ in the organization from multi-skilling itself,
it may even be able to compensate somewhat for non-innovative management,
always providing that the internal information gatekeeping is both open and open
to accepting change.
Please note that the positive effects of exogenous networks does not immediately
translate into having large exogenous networks, the innovation arriving has to have
immediate applicability, implying that external networks need to be highly relevant
rather than simply extensive.
Intellectual Property and the concept of knowledge trails.
For the past thirty years the concept of “relentless innovation” has gained wide
currency. Despite this and especially in the SME sector, the majority of companies
are reluctant learners, happy to exist, to be making a profit, going for moderately
safe goals in inherently unstable Porterian “weather” and unwilling to disturb the
balance with risky changes. This is especially true for the majority as they age,
become stable and thus become “the incumbents”. They absorb data from the
environment and process it according to pre-set routines. These routines can be
represented in simple terms as a “preferred relationships” or a Knowledge Trail.
Knowledge trails do not themselves represent an innovation, but they do represent a
cost-saving and they are therefore accorded value, the extent of which is dependent
upon where they are – high or low – in the knowledge valley fold they inhabit. A
mundane knowledge trail could be e.g. customer complaints, where any complaint is
processed according to a well-known procedure and employees do not start from
scratch each time a customer communication arrives.
Knowledge trails are also associated with intellectual property. A relatively high value
knowledge trail could be e.g. the fairly routine registration of trade-marks, design or
database rights etc. Clearly of even higher value would be an example concerned
with the invention of a new process or device then writing and filing for patent
protection. Interestingly the value and trajectory of a knowledge trail is fairly
constant, even when the value of the company falls. This is illustrated in figure 7.
Figure 7: The knowledge valley fold showing (red) a meandering low-level
knowledge trail and in contrast a knowledge trail (blue) that is facilitated by the
“bridge” of a patent (or other intellectual property).
Thus it becomes apparent:
12. Patents (and proportionally other intellectual property) enable knowledge trails
to firstly retain value but also to bridge any gap in organizational profitability
during growth which in turn enables the respective knowledge trail to start
again at an over-proportionally (compared to the organization) high valuation
as support returns.
This explains the popularity of patenting and intellectual property amongst aspiring
high-innovation start-ups in e.g. biotechnology because it retains value also when
the organization may otherwise be struggling.
Conclusion
One should not confuse conclusions drawn from observation with those derived from
scientific experimentation. Less than ten years of in silico experimentation has
explained rationally countless observations made in the field for over fifty years. Not
only are these techniques one or two orders of magnitude faster than conventional
methods but they are also rather simple to apply. Throughout this communication
twelve points have been made which the practitioner can easily appreciate.
Academics and practitioners alike can now pinpoint their clients’ position on the
knowledge valley fold and for the first time are able to advise clients in quantitative
terms the best way forward and pitfalls to avoid.
References
Chib, S. and Greenberg, E. (1996): Markov Chain Monte Carlo simulation methods
in econometrics. Econometric Theory, 12 (3), 409-431. Available at
http://econwpa.repec.org/eps/em/papers/9408/9408001.pdf Last accessed March
2016.
Desouza, K.C. and Awazu, Y.A. (2006): Knowledge management at SMEs: five
peculiarities. Journal of Knowledge Management, 10 (1), pp 32 - 43.
Ditillo, A. (2004): Dealing with uncertainty in knowledge-intensive firms: the role of
management control systems as knowledge integration mechanisms. Accounting,
Organizations and Society, 29, pp 401–421.
Greiner, L. (1972). Evolution and revolution as organisations grow. Harvard Business
Review, 50, July/august, 37-46.
Mellor, R.B. (2005) Sources and Spread of Innovation in Small E-commerce
Companies, Copenhagen, Globe.
Mellor, R.B. (2011) Knowledge Management and Information Systems, Strategy for
Growing Organizations, London, Palgrave Macmillan, ISBN 9780230280434.
Mellor, R.B. (2014a) Knowledge valley theory, International Journal of Knowledge-
Based Development, Vol. 5, No. 4, pp.5–16.
Mellor, R.B. (2014b) The use of knowledge assets: modelling the potential effect of
adding innovators to low-innovation and high-innovation SMEs, International Journal
of Knowledge-Based Development, Vol. 5, No. 4, pp.367–380.
Mellor, R.B. (2015a) Computer-Modelling the Innovation-Based Theory of the Firm.
The proceedings of the European Conference on Knowledge Management (ECKM-
15) at Udine, Italy, 2-5 September 2015. ISBN: 9781910810460, pp. 532-540.
Mellor, R.B. (2015b) Modelling the value of external networks for knowledge
realisation, innovation, organisational development and efficiency in SMEs.
International Journal of Knowledge-Based Development, Vol. 6, No. 1, pp. 3-14.
Robinson, S. (2014): Simulation; the Practice of Model Development and Use.
London. Palgrave Macmillan.
Serenko, A., Bontis, N. and Hardie, T. (2007): Organizational size and knowledge
flow, a proposed theoretical link. Journal of Intellectual Capital, 8 (4), pp 610-627.
Tsai, M-T. and Huang, Y-C. (2008): Exploratory learning and new product
performance: The moderating role of cognitive skills and environmental uncertainty.
The Journal of High Technology Management Research, 19, pp 83-93.
Williamson, O. E. (1985): The economic institutions of capitalism. The Free Press,
New York.
Williamson, O.E. (1993): Opportunism and its critics, in: Managerial and decision
economics, 14, pp 97-107.
Top Related